Deep learning of data-driven Heath-Jarrow-Morton models
We consider a data-driven version of Heath-Jarrow-Morton models in the context of interest rate modeling. More precisely, we use as driving process a vector of forward rates for a set of representative maturities whose characteristics can be easily estimated from market data. We then parametrize the volatility function via neural networks, thus considering the framework of neural SPDEs. Their parameters are then learned by calibrating the model to past market yield curves. This results in a data-driven arbitrage-free generation/prediction of yield curves relevant for risk management purposes. Our setup also allows for the incorporation of stochastic discontinuities, a key feature in interest rate markets, which can for instance be seen from the jumps of the term structures in correspondence to monetary policy meetings of the European Central Bank. We illustrate our deep learning procedure by reconstructing and forecasting the Euro area yield curves. The talk is based on joint work with Claudio Fontana and Alessandro Gnoatto.
Area: CS58 - Recent advances in Heath-Jarrow-Morton modelling in finance (Claudio Fontana and Alessandro Gnoatto)
Keywords: Neural SPDEs, HJM-models, data-driven arbitrage-free interest rate generation
Please Login in order to download this file